Fixes for webmerc projection (fixes #1078)

[rouault]

This is intended to supersed #1080
with a number of differences.

What is kept from #1080 is not forcing the ellipsoid_params to be the one
of a sphere. This is not required for correct coordinate computation and
avoid lying on the various distorsion parameters.

For better interoperability with EPSG, we also no longer force the
lam0 parameter to 0, because
https://www.epsg-registry.org/export.htm?gml=urn:ogc:def:method:EPSG::1024
has a provision for it, even if in practice they will always be zero
phi0 should always be zero and is not used by the formulas.

Another difference with the #1080 approach is that we do not force the
WGS84 ellipsoid. Perhaps someone will use webmerc for another planet,
even if that is a crazy idea...

[kbevers]

Disclaimer: I haven't looked too closely at the code.

Does this solve the original issue with errornoues projection factors as reported in #1078?

[rouault]

Does this solve the original issue with errornoues projection factors as reported in #1078?

yes. I've checked that the values proj -v now outputs for the Meridian scale, Parallel Scale and Angular Distorsion are the ones given by EPSG Guidance 7.2 page 44 for the test point they give

$ ./src/proj +proj=webmerc +ellps=WGS84 -V
#Web Mercator / Pseudo Mercator
#	Cyl, Ell
#	
# +proj=webmerc +ellps=WGS84
#Final Earth figure: ellipsoid
#  Major axis (a): 6378137.000
#  1/flattening: 298.257224
#  squared eccentricity: 0.006694379990
-100.33333332372804 24.381786961614658
Longitude: 100d20'W [ -100.33333332 ]
Latitude:  24d22'54.433"N [ 24.381786962 ]
Easting (x):   -11169055.58
Northing (y):  2800000.01
Meridian scale (h) : 1.10342637  ( 10.34 % error )
Parallel scale (k) : 1.09729143  ( 9.729 % error )
Areal scale (s):     1.21078030  ( 21.08 % error )
Angular distortion (w): 0.319
Meridian/Parallel angle: 90.00000
Convergence : 0d [ -0.00000000 ]
Max-min (Tissot axis a-b) scale error: 1.10343 1.09729

[kbevers]

Awesome! I am happy with this PR.